Takano Fumiki, Hiratsuka Masaki, Takahashi Kazuaki Z
Kogakuin University, 1-24-2 Nishi-Shinjuku, Tokyo, 163-8677, Japan.
National Institute of Advanced Industrial Science and Technology (AIST), Research Center for Computational Design of Advanced Functional Materials, Central 2, 1-1-1 Umezono, Tsukuba, Ibaraki, 305-8568, Japan.
Sci Rep. 2024 Oct 13;14(1):23908. doi: 10.1038/s41598-024-74525-y.
The microphase-separated structures of block copolymers are inherently highly ordered local structures, commonly characterized by differences in domain width and curvature. By focusing on diblock copolymers, we propose local order parameters (LOPs) that accurately distinguish between adjacent microphase-separated structures on the phase diagram. We used the Molecular Assembly structure Learning package for Identifying Order parameters (MALIO) to evaluate the structure classification performance of 186 candidate LOPs. MALIO calculates the numerical values of all candidate LOPs for the input microphase-separated structures to create a dataset, and then performs supervised machine learning to select the best LOPs quickly and systematically. We evaluated the robustness of the selected LOPs in terms of classification accuracy against variations in miscibility and fraction of block. The minimum local area size required for LOPs to achieve their classification performances is closely related to the characteristic sizes of the microphase-separated structures. The proposed LOPs are potentially applicable over a large area on the phase diagram.
嵌段共聚物的微相分离结构本质上是高度有序的局部结构，通常以畴宽度和曲率的差异为特征。通过聚焦于二嵌段共聚物，我们提出了局部序参量（LOPs），其能够准确区分相图上相邻的微相分离结构。我们使用用于识别序参量的分子组装结构学习软件包（MALIO）来评估186个候选LOP的结构分类性能。MALIO为输入的微相分离结构计算所有候选LOP的数值以创建一个数据集，然后进行监督机器学习以快速且系统地选择最佳的LOP。我们根据分类准确率评估所选LOP在混溶性和嵌段分数变化方面的稳健性。LOP实现其分类性能所需的最小局部区域大小与微相分离结构的特征尺寸密切相关。所提出的LOP在相图上的大面积范围内具有潜在的适用性。
